Post on 20-Jul-2022
NBER WORKING PAPER SERIES
HEADSTRONG GIRLS AND DEPENDENT BOYS:GENDER DIFFERENCES IN THE LABOR MARKET RETURNS TO CHILD BEHAVIOR
Robert KaestnerOfer Malamud
Working Paper 29509http://www.nber.org/papers/w29509
NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts Avenue
Cambridge, MA 02138November 2021
We would like to thank Alice Eagly, David Figlio, Seema Jayachandran, Kevin Lang, and Terri Sabolfor their valuable comments. Timea Viragh, Yuheng Wang, and Yaling Xu provided excellent researchassistance. All errors are our own. The views expressed herein are those of the authors and do notnecessarily reflect the views of the National Bureau of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies officialNBER publications.
© 2021 by Robert Kaestner and Ofer Malamud. All rights reserved. Short sections of text, not to exceedtwo paragraphs, may be quoted without explicit permission provided that full credit, including © notice,is given to the source.
Headstrong Girls and Dependent Boys: Gender Differences in the Labor Market Returns toChild BehaviorRobert Kaestner and Ofer MalamudNBER Working Paper No. 29509November 2021JEL No. J16,J24
ABSTRACT
The authors use data from the Children of the National Longitudinal Survey of Youth (C-NLSY79)to examine gender differences in the associations between child behavioral problems and early adultearnings. They find large and significant earnings penalties for women who exhibited more headstrongbehavior and for men who exhibited more dependent behavior as children. In contrast, there are nopenalties for men who were headstrong or for women who were dependent. While other child behavioralproblems are also associated with labor market earnings, their associations are not significantly differentby gender. The gender differences in headstrong and dependent behavior are not explained by education,marriage, depression, self-esteem, health, or adult personality traits. However, one potential explanationis that these gender differences are a consequence of deviations from gender norms and stereotypesin the workplace.
Robert KaestnerHarris School of Public PolicyUniversity of Chicago1307 East 60th Street (Room 3057)Chicago, IL 60637and NBERkaestner@uchicago.edu
Ofer MalamudSchool of Education and Social PolicyNorthwestern UniversityAnnenberg Hall2120 Campus DriveEvanston, IL 60208and NBERofer.malamud@northwestern.edu
2
A substantial literature has documented the significant relationship between cognitive
skills, measured in childhood and adolescence, and adult earnings (e.g., Murnane, Willet and
Levy, 1995). More recently, studies have also reported significant associations between adult
earnings and childhood “non-cognitive skills”, such as socio-emotional behaviors and
temperament (e.g., Heckman, Stixrud and Urzua, 2006, and Lindqvist and Vestman, 2011).
However, there has been relatively little research on gender differences in the labor market
returns to childhood “non-cognitive skills”, especially when disaggregated into specific domains
of behavior. If men and women are paid differently for the same early-life skills or behaviors,
this may contribute to gender gaps in earnings and suggest the presence of labor market
discrimination (Blau and Kahn, 2017). Gender differences in the labor market returns to
childhood behaviors can also provide insights about the role of gender norms and stereotypes in
the labor market.
We use data from the Children of the National Longitudinal Survey of Youth 1979 (C-
NLSY79) survey to examine associations between several distinct child behavioral problems
(ages 4 to 12) and early adult earnings (ages 24 to 30). Our measures of child behavioral
problems are drawn from an abbreviated version of the Child Behavior Checklist (CBCL) that is
reported by parents. They were “…designed to measure some of the more common syndromes of
problem behavior found in children” (Zill, 1985). The CBCL is one of the most widely used
assessments of children’s emotional and behavioral problems, and shown to have strong
predictive validity and to be measurement invariant with respect to child gender. Like the CBCL
on which they are based, the measures of behavioral problems in the C-NLSY79 have also been
shown to have strong predictive validity (Parcel and Menaghan, 1988). The use of these
3
disaggregated categories of child behavior is novel and empirically important because broader
constructs can obscure the heterogeneous relationships of individual behaviors with earnings.
While our study is unique in examining gender differences in the relationship between
child behavior and labor market outcomes, there is a large literature that examines the
relationship between adult personality and labor market outcomes, as well as the gender
differences in these relationships.1 However, it is important to distinguish between the labor
market returns to behaviors observed in childhood vs. self-reported personality traits measured in
adulthood. First, these are conceptually different psychological constructs, and as we show later,
they are not very highly correlated. Second, our measures are not subject to the biases of self-
reports. Third, and perhaps most important, is the possibility of reverse causality in analyses that
correlate contemporaneous measures of earnings and personality. Our information on child
behavior was collected at an early age, long before labor market entry and before schooling is
completed. We also present some evidence suggesting that our estimates are not confounded by
other background characteristics. However, since we cannot rule this out altogether, we are
careful throughout this article to refer to the relationships between child behaviors and adult
outcomes as associations
We also explore whether gender differences in the returns to child behavioral problems
are mediated by educational attainment, health, marriage, children, as well as adult personality.
Through these analyses, we also add to an existing literature that examines the relationship
between child behavior problems and educational attainment.2 Furthermore, our study
contributes to the developmental psychology literature focused on the relationship between
1 See Mueller and Plug, 2006; Heinek and Anger, 2010; Cobb-Clark and Tan, 2011; Heinek, 2011; Judge, Livingston, and Hurst, 2012; Nyhus and Pons, 2012; Fletcher, 2013; Gensowski 2018 2 See Hinshaw, 1992; Segal, 2008; Diprete and Jennings 2011; Kristoffersen et al. 2015; Owens, 2016; Papageorge, Ronda, and Zheng, 2019
4
childhood behavior and adult personality (Caspi et al. 2005; Donnellan et al. 2015) by estimating
these associations for a large, recent cohort of US children that, to our knowledge, has not been
previously examined. Finally, our research is also related to the effects of attention-
deficit/hyperactive disorder (ADHD) on educational attainment and labor market outcomes
(Currie and Stabile, 2006; Fletcher and Wolfe, 2008, 2009; Fletcher, 2014).
The remainder of the article proceeds as follows. The next section describes our measures
of child behavior and summarizes the gender differences in behavior for our sample. We then
report our estimates of the associations between earnings and these behaviors by gender.
Following that, we consider the associations of child behavior with other outcomes and explore
potential mechanisms which may explain our results. Finally, we summarize our findings and
discuss the limitations of our analyses and avenues for further research.
Background
Measures of child behavioral problems in the C-NLSY79
The measures of child behavioral problems available in the C-NLSY79 are derived from
an abbreviated version of the Child Behavior Checklist (CBCL) that is one of the most widely
used assessment of children’s emotional and behavior problems. The items that comprise these
measures “were selected because they were not too rare in the general child population; had a
demonstrated ability to discriminate children who had received clinical treatment from those who
had not; and tapped some of the more common behavior syndromes in young people” (Zill,
1990). The full set of items in this abbreviated version of the CBCL are commonly referred to as
the Behavior Problems Index (BPI). However, in accordance with the CBCL, they were also
designed to identify specific behavioral syndromes in children.
5
The specific behavioral syndromes identified by Zill (1985, 1990) and recorded in the C-
NLSY79 are referred to as (a) antisocial behavior, (b) anxiety/depressed mood, (c) headstrong
behavior, (d) hyperactive behavior, (e) dependent behavior, and (f) peer conflict. Appendix Table
1 lists the questions that comprise these distinct behavior problem subscales along with their
associated syndromes. We also describe the items used to construct the headstrong and
dependent subscales below because, as we show later, they are the ones associated with
significant gender differences in earnings. Children score high on the headstrong subscale when
their caregiver reports that he or she (i) argues too much, (ii) has strong temper and loses it easily
(iii) is disobedient at home, (iv) is stubborn, sullen, or irritable, and (v) is rather high strung,
tense, and nervous. Children score high on the dependent subscale when their mother reports that
he or she (i) demands a lot of attention, (ii) clings to adults, (iii) cries too much, and (iv) is too
dependent on others.
The behavioral problem measures in the C-NLSY79 are based on maternal reports of
children between the ages of 4 and 12. The response categories for these questions are “often
true,” “sometimes true,” and “not true”. The scores for the behavioral syndromes are produced
by summing across the dichotomized responses to the relevant subset of items (whereby each
item answered “often” or “sometimes true” is given a value of one). We average the scores
across all surveys for which the child is present between ages 4 and 12 in order to improve
reliability.3 However, our findings are unchanged if we use measures of behavioral problems
assessed at early ages (4-6) or later ages (8-12), which is consistent with the fairly stable pattern
3 Approximately 86% of the sample of children were present for 4 or 5 survey waves, which represent the maximum number of possible surveys to be present given that surveys were fielded every two years. We assessed whether controlling for the number of surveys present affected estimates and there was virtually no change in our results.
6
of gender differences in these child behaviors between ages 4 and 12 (see Appendix Figure 1,
where blue lines represent boys and orange lines represent girls).
The validity of behavioral problem measures
The behavioral problem measures in the C-NLSY79 have been extensively analyzed and
validated across different populations. Zill (1985) uses principal components analysis to verify
that the clusters of items designed to identify these behavior syndromes represented separable
dimensions with common underlying factors using the 1981 Child Health Supplement to the
National Health Interview Survey (NHIS). Parcel and Menaghan (1988) correlate the behavioral
problem measures with a variety of social and demographic variables in the C-NLSY79 to
provide evidence of construct validity. The CBCL, which is the primary source for the
behavioral measures used in the study, has been subject to even more extensive validation to
confirm that it identifies specific behavioral syndromes and discriminates between children
referred and non-referred to treatment across different populations (e.g. Edlebrock and Costello
1988; Jensen et al. 1993; Chen et al. 1994; Drotar et al. 1995; Kasius et al. 1997; Greenbaum et
al. 2004; Seligmman et al. 2004; Ivanova et al. 2007; Ferdinand 2008; Nakumra et al. 2009;
Ebustani et al. 2010; and Ivanova et al. 2019).
One issue that merits discussion is the possibility of gender bias in mothers’ assessments
of child behavioral problems. Using the full set of items in the CBCL, Konold et al. (2004) show
that the general form and factor loadings of child behavior problems are measurement invariant
(i.e., unbiased) with respect to the child’s gender and across mother’s and father’s assessments.
The measurement invariance by child gender is further confirmed by van der Sluis (2017) for a
large clinical sample.
7
Despite the evidence that the CBCL is measurement invariant with respect to gender, we
also assessed whether it is a potential issue for our sample. Accordingly, we re-estimated the
models for our primary outcome of earnings using gender-specific, standardized measures of
behaviors, as suggested by van der Slius (2017) and others. Results from this analysis were
virtually identical to those reported below that uses the common measures. We also examined
whether one particular question in the dependent subscale—does the child cry too much—had a
particularly strong influence on results related to the dependent subscale.4 It did not. Finally, we
also address this issue by showing that our results are robust to controlling for an extensive set of
mothers’ characteristics. Overall, there is substantial evidence that the behavioral subscales are
measuring the same constructs for males and females, and that the results reported below are not
an artifact of measurement problems.
Gender differences in behavior, skills, and other covariates
Differences in child behavior and skills by gender are well known and have been
described in prior work (McLeod and Kaiser, 2004; Else-Quest, et al. 2006; Halpern et al. 2007;
Duncan and Magnuson, 2011; Owens 2016; Autor et al. 2019). However, we briefly describe the
gender differences for our sample of children who were born between 1981 and 1990 and
observed as adults between the ages of 24 to 30 (i.e., between 2006 and 2014).5 As shown in
Table 1, boys have significantly higher rates of hyperactive, anti-social, headstrong and peer
4 This question about crying too much has been mentioned as one that may not be measurement invariant by child gender (van der Slius, 2017). Indeed, Yarnell et al. (2013) reported that the question cries too much was not measurement invariant in a sample of children from Mauritius. To assess whether this specific component is a problem in our sample, we reconstructed the dependent subscale omitting this question and re-estimated the models reported below. Our results were essentially unchanged. 5 The sample consists of 3,477 unique children whom we observe 7283 times as adults: 1,052 are observed once; 1,163 are observed 2 times; 1,143 are observed 3 times; and 119 are observed 4 times.
8
conflict behavior while girls have significantly higher rates of dependent behavior. The gender
differences in behavior are quite similar over childhood years, as seen Appendix Figure 1.
The C-NLSY79 measured math and reading skills between the ages of 5 and 13 using the
Peabody Individual Achievement Test (PIAT) in Mathematics, Reading Recognition, and
Reading Comprehension. Table 1 indicates that math scores are almost identical for boys and
girls while reading scores are about 0.1 standard deviations higher for girls than boys.
Finally, the C-NLSY79 also provides extensive information about the child’s family
background characteristics, including mother’s education, her Armed Forces Qualification Test
(AFQT) score, marital status at birth, self-esteem, internal-external locus of control and
indicators for whether the household had magazines, newspapers and a library card. As shown in
Table 1, most of these background characteristics are not substantially different by gender,
although there are small differences in mother’s marital status and age at birth, and whether the
household had a library card. Additional evidence related to selection on family background
factors is provided in Appendix Table 2. This table presents regression estimates of the gender
differences in the relationship between several maternal characteristics and our six child
behaviors of interest. Of the 36 estimates shown, only three differ significantly by gender.
Labor market returns to child behavior
Earnings are our primary outcome and defined as self-reported annual earnings in the
previous year.6 The goal of our study is to estimate the associations between several distinct
child behavioral problems and adult earnings, and to assess whether there are gender differences
6 The sample consists of those with valid earnings including zeros. Earnings information is missing (i.e., with an invalid response) for approximately 11% of the sample. None of the child behaviors are significantly related to the probability of being in the sample.
9
in those associations. We estimate a series of OLS regression models to obtain these estimates.
As described in the previous section, we focus on continuous measures of child behavior based
on dichotomized responses to the underlying items, as constructed in the C-NLSY79. However,
we standardize these measures to have mean 0 and a standard deviation of 1 to facilitate
comparisons of the magnitudes of the coefficients.
All of our regression models include controls for child race/ethnicity (non-Hispanic
white, non-Hispanic Black, Hispanic), birth order, each year of age at time of interview, and each
birth year. Thus, we compare children assessed at similar ages and from similar birth cohorts.
We also include math and reading achievement test scores and controls for mother characteristics
to address potential confounding factors and differences in reporting, allowing for differential
effects of all variables by gender.7 The standard errors of the regression estimates allow for non-
independence between multiple observations per person.
Table 2 presents results from these regressions. The first two columns show estimates for
the full sample of men and women respectively. We observe large and significant earnings
penalties for men who exhibited more dependent behavior in childhood. A one standard
deviation (1σ) increase in dependent behavior is associated with a $1632, or 6 percent, decline in
earnings. In contrast, there are no penalties for women characterized as dependent; indeed, the
coefficient is positive, albeit small ($431) and statistically insignificant. We also observe large
and significant earnings penalties for women who exhibited more headstrong behavior in
childhood. A 1σ increase in headstrong behavior is associated with a $2092, or 10 percent,
decline in earnings. There are no penalties for men characterized as headstrong; again, the
7 Mother’s covariates include dummy variables for each age at birth; dummy variables for education (LTHS, HS some college, BA or more); AFQT score and its square; marital status at birth (married, never married, other); dummy variables for quartile of self-esteem scale; dummy variables for quartile of Rotter scale; and dummy variables indicating whether household of child had magazines, newspapers and a library card.
10
coefficient is positive, but small ($339) and insignificant. The third column, which is effectively
the difference of columns one and two, confirms that the association of dependent and
headstrong behavior with earnings is significantly different between males and females.
In the next three columns, we confirm that our results are robust to an alternative
specification using the natural logarithm (log) of earnings in the sample of working men and
women (i.e., those who worked in the last year and had non-zero earnings). Estimates indicate
that a 1σ increase in dependent behavior is associated with an 8.5 percent decline in earnings for
men; the corresponding increase of 3.6 percent earnings for women is economically meaningful,
but not statistically significant. We also observe that a 1σ increase in headstrong behavior is
associated with a 9.9 percent decline in earnings for women; the increase of 3.5 percent earnings
for women is again economically meaningful, but statistically insignificant. The final column
shows that the labor market returns to being dependent or headstrong are significantly different
by gender. The similar pattern of estimates in analyses of earnings using the full sample
including those not working and for the sample of those working with positive (log) earnings
implies that child behavior problems are not strongly related to employment.
While other child behaviors are also associated with labor market earnings, their
associations are generally weaker, less consistent across models, and not significantly different
by gender. For example, both men and women suffer earnings penalties for being relatively more
hyperactive, anti-social and having more peer conflict, but the gender differences in these
associations are not significant and much smaller than those for dependent and headstrong
behavior. Similarly, both men and women who are relatively more anxious/depressed as children
have higher earnings, but the gender difference is not significant. Note that we did not find
significant gender differences in the returns to the overall measure of child behavior problems
11
(i.e., the BPI), or for externalizing and internalizing behavior. As noted earlier, these results
suggest that using the overall measure of behavior problems obscures the effects associated with
specific behavior syndromes.
The gender differences in Table 2 are robust to alternative regression specifications, as
shown in Appendix Table 3.8 First, estimates are similar if we restrict the effects of
mother/family characteristics on earnings to be the same by gender. This finding, together with
the fact that there are few differences in mother/family characteristics by gender, suggests that
the estimates in Table 2 are unlikely to be confounded by omitted mother/family factors, or by
biases related to maternal reporting of behaviors. The estimates are also qualitatively similar
when restricted to earlier (born 1981-84) and later (born 1985-90) cohorts of children, although
we lose precision because of the smaller sample sizes. Our results also hold if restrict the sample
to observations of earnings from 2010 onwards, after the initial years of the subprime crisis.
Finally, our main findings remain unchanged when we include only dependent and headstrong
measures omitting other behavioral measures, and when we omit mother and family background
characteristics.9
We also consider the possibility of multiple testing bias for our “family” of hypotheses
related to the gender differences in associations between earnings and our six child behaviors of
interest. To the extent that our analysis is exploratory—it is the first to examine gender
differences the associations between earnings and these behaviors—it is not clear whether it is
necessary to adjust for multiple testing bias because doing so raises the likelihood of a false
8 Our findings are also robust to using continuous measures of math and reading achievement, excluding measures of math and reading achievement, and using categorical measures of child behaviors instead of continuous measures. 9 We also examined the effects on low- and high-earnings defined using the 25th and 75th percentiles of pooled (across genders) earnings. Results from these analyses are consistent with the results with our main analysis of earnings suggesting that the gender differences are apparent throughout the earnings distribution.
12
negative finding (Wason et al. 2014; Guowei et al. 2017). Moreover, the behaviors we study are
distinct, and a finding about the association between earnings and one behavior does not have
any bearing on the inference and conclusions regarding the relationship between earnings and
another behavior (Proschan and Waclawiw 2000).10
Nevertheless, to provide some evidence on this issue of multiple testing bias, we adjust
the level of significance of our six estimates of gender differences in associations between
earnings and child behaviors using the Holm-Bonferroni correction. The p-value associated with
the estimate of the gender difference related to dependent behavior ($2063) is 0.014, which after
adjustment for six hypotheses (i.e., multiplying p-value by six) is 0.084. While not significant at
the conventional 0.05 level, it is close and incorporates a severe penalty for the possibility of any
type I error. A similar adjustment applied to the p-value of the estimate for the gender related to
headstrong behavior is 0.19. Moreover, these adjustments yield even lower p-values when
applied to the estimates from the log earnings specification. Thus, even if we adopt a stringent
adjustment for multiple comparison bias, such as the Holm-Bonferroni, the significance levels of
estimates of gender differences in associations between earnings and dependent and headstrong
behaviors are not far from conventional levels of significance.
Potential mechanisms
What explains the gender differences in the association of child behavior with early adult
earnings? To address this question, we first examine the gender differences in the associations of
child behaviors with potentially mediating factors that have been shown to affect earnings such
10 In contrast, if we were interested in inferring whether childhood behaviors broadly construed were associated with earnings, then it would be more important to adjust for multiple testing bias.
13
as employment, educational attainment, marriage, fertility, health, and personality.11 Second, we
examine whether our estimates of the associations of child behaviors and earnings are moderated
by factors that determine workplace settings and may, in turn, relate to prejudice in the
workplace.
Mediating Factors
Table 3 presents gender differences in the associations of child behaviors with
employment, educational attainment, marriage, fertility, health using the analogous regression
models and the same sample as used in Table 2 (the corresponding associations by gender are
relegated to Appendix Table 4). The first two columns examine whether these gender differences
might be driven by employment and work hours. While the gender difference in the associations
of headstrong behavior with employment is marginally significant, driven by a 3.6 percentage
point reduction in employment for headstrong women shown in Appendix Table 4, it is not
sufficient to explain the large earnings differences we observed in Table 2. This fact is evidenced
by the estimates of gender differences for headstrong behavior and (log) earnings, which is
limited to those working. Moreover, there are no significant gender differences in the association
of work hours with either headstrong or dependent behavior.
The next two columns of Table 3 examine educational attainment, measured as years of
completed education and attainment of a bachelor’s degree. Here, the gender difference in the
association of dependent behavior with years of schooling is marginally significant, but again
11 There are extensive literatures documenting the independent effect of education, marriage, and health on earnings. See, e.g., Card (1999) for education, Ginther and Zavodny (2001) for marriage, and Prinz et al. (2018) for health. We chose to explore potential mechanisms in this manner rather than adding mediators to the earnings regression model because it avoids the “bad control” problem.
14
insufficient to explain the large earnings difference we observed in Table 2.12 Moreover, there
are not significant gender differences in the association of college completion with either
headstrong or dependent behavior. Neither do the following two columns of Table 3 reveal any
significant associations between child behavior and the likelihood of marriage or number of
children.
The final two columns of Table 3 show large gender differences in the associations of
dependent and headstrong behavior with the self-reported likelihood of being in poor health.
Indeed, the patterns in Appendix Table 4 indicate that men are 3.2 percentage points more likely
to report that they are in poor health with a 1σ increase in dependent behavior. However, it seems
unlikely that health represents a potential pathway for explaining the gender differences in how
child behavior affects earnings because there are relatively few people in poor health and health
often affects earnings through employment (on which we find little effect). Furthermore, we do
not observe any significant differences by gender in the likelihood of being in good health.
We also examine the associations of child behavior with measures of depression, self-
esteem, mastery, and the Big-5 personality traits in early adulthood.13 Table 4 presents estimates
for the gender differences in the associations of child behavior and scores on these psychological
outcomes (with the main effects provided in Appendix Table 5). For our research question, the
most important finding revealed by the estimates in Table 4 is the relatively few significant
gender differences (only 3 out of a possible 48). Thus, the gender differences in the associations
12 Assuming a 10% increase in earnings for each year of schooling, the gender difference in schooling between dependent women and men would only result in a 1.5% difference in earnings (approximately one fifth of the gender difference in earnings in Table 2). 13 Depression is measured using the Center for Epidemiologic Studies Depression (CES-D) scale; mastery is measured using the Pearlin Mastery scale. These variables are measured between ages 18 and 23 in the C-NLSY.
15
between childhood behavior and earnings do not appear to be mediated by depression, self-
esteem, or adult personality.
It is notable that the estimates in Appendix Table 5 do not reveal many strong
associations between child behaviors and adult personality traits. This finding contributes to the
developmental psychology literature that studies the life course linkages between child
temperament and adult personality (e.g., Caspi et al. 2005; Shiner and DeYoung 2013). In
contrast to some previous findings, our novel estimates from a large and recent U.S. cohort of
children indicate relatively little correlation between child behavioral problems (at ages 4 to 12)
and personality traits (at ages 24 to 30).
Finally, in Table 5, we examine the mediating role of occupation (at the 3-digit level)
directly, which is a variable that does not lend itself to an analysis similar to that in Table 2. This
analysis explores whether occupational sorting mediates the gender differences in associations
between headstrong and dependent behavior and earnings. Estimates in Table 5 indicate that
adding controls for education and occupation does not appear to affect the estimates of gender
differences in the returns to headstrong behavior, although it does reduce the magnitude of the
gender differences in the returns to dependent behavior by 30-40 percent. Including other
mediators has almost no additional effect. We recognize that any analysis which conditions on
education and occupation and other variables reflecting personal choices may introduce bias if
these variables are also affected by child behavior differentially by gender or are influenced by
earnings (reverse causality). This potential problem also applies when these variables are used as
dependent variables, as in Tables 3 and 4. Nevertheless, it does not appear to be the case that our
results are explained by occupation or by differences in education and other choices. In other
16
words, to the extent that there are these gender-specific associations between behavior and
earnings, it is happening within occupation and within educational levels.14
Moderating Factors
The absence of any evidence that gender differences in the associations between child
behaviors and earnings are mediated by a relatively large set of variables known to affect
earnings leaves open the question of the mechanism underlying this finding. If the character
traits that underlie these child behaviors persist into adulthood, one potential explanation for the
gender differences in the associations of dependent and headstrong behavior with earnings is
prejudice in the workplace.15 For example, if children who exhibited behaviors that deviated
from gender norms and stereotypes are subsequently penalized in the labor market. We attempt
to assess this hypothesis by examining whether gender differences in the labor market returns to
dependent and headstrong behaviors differ by education (less vs. more than high school) and
occupation (blue-collar vs. white-collar), which are variables that reflect distinct workplace
settings and may be related to different views about how behavioral traits are gender-specific.16
We also examine whether the gender differences we observe depend on whether individuals are
employed in occupations that are female-dominated or not (i.e. if the fraction of women is higher
than 75 percent).
14 We also examined whether child behaviors are related to occupational “work styles” as reported in the O*NET database (such as “persistence”, “concern for others”, and “independence”). We linked this information to occupational indicators in the CNLSY and estimated regression models identical to those in Table 2 using the scores for the importance of occupational worker styles as dependent variables. The vast majority of these coefficients, reported in Appendix Table 6, are insignificant, suggesting that there is very little sorting of individuals to occupations based on these work styles, and essentially no differential sorting by gender. 15 Presumably, these character traits are not captured by the measures of adult personality available in the CNLSY (given that we do not find a strong correlation between child behaviors and these measures). 16 For example, data from Pew opinion polls in 2008 and 2014, as well as a GfK poll conducted for Eagly et al (2019), indicate that people with more than a high school education are relatively less likely to associate men with emotional/sensitive traits or women with decisive/stubborn traits as compared to less individuals with less education.
17
Table 6 presents estimates of the associations of child behaviors with earnings stratified
by education, blue vs. white-collar occupations, and female-dominated occupations. The gender
differences in labor market returns to dependent and headstrong behavior are large and
significant for those with less than a high school education, while substantially smaller among
more educated individuals. A similar pattern for the associations between child behaviors and
earnings emerges when we stratify by whether a person works in a blue-collar occupation or not:
the gender differences in labor market returns to dependent and headstrong behavior are large
and significant for those working in a blue-collar occupation, while substantially smaller among
those in non-blue-collar jobs.17 These results provide suggestive evidence that workplace
characteristics (such as workplace prejudice) may be related to our findings of gender differences
in the associations of dependent and headstrong behavior with earnings. Nevertheless, gender
differences do not appear qualitatively different when stratified by whether or not occupations
are female-dominated.
Summary
To summarize our main findings, we observe large and significant earnings penalties for
women who exhibited more headstrong behavior, and for men who exhibited more dependent
behavior, as children. There are no significant or economically meaningful penalties for men
who were headstrong or for women who were dependent. While other child behavioral problems
are also associated with labor market earnings, their associations are not significantly different
by gender. These results are robust to alternative specifications of earnings (e.g., levels or logs)
17 Interestingly, when estimating the returns to child behavior separately by gender, the earnings penalty for women characterized as headstrong is substantially larger, and only significant among the more educated or those who work in non-blue-collar occupations. The earnings penalty for men characterized as dependent is substantially larger, and only significant, among the less educated and those who work in blue-collar occupations.
18
and child behaviors (e.g., linear or categorical). The results are also similar with and without
adjustment for differences in academic achievement or family background, and when allowing
for the effect of family background to differ by gender. While we have been careful to avoid
making strong causal claims about the estimated relationships, these sensitivity analyses do
assess whether some of the more important potentially confounding influences due to family
background are present. Results suggest they are not.18
The differential returns to dependent and headstrong behavior by gender are not mediated
by significant gender differences in the associations between child behaviors and the likelihood
of employment, work hours, marriage, fertility, or self-esteem. Nor do we find gender
differences in the associations between child behaviors and adult personality traits, mental health
(CESD) or the probability of being in good health. While we do find that men who were
characterized as dependent are significantly more likely to report being in poor health, and this
association is significantly different from the corresponding one for women, these differentials
are too small for health to be a significant mediator of the gender difference in the association
between dependent behavior and earnings.
Notably, we find heterogeneous gender differences in the returns to headstrong and
dependent behavior by education and occupation. This heterogeneity is suggestive of the role of
workplace settings, and perhaps workplace prejudice, in explaining gender differences between
these child behaviors and early adult earnings. Thus, one potential explanation of our findings is
18 Unfortunately, we don’t believe there are any feasible quasi-experimental approaches that can isolate exogenous variation in these specific child behaviors, similar to other related studies such as Attanasio, Blundell, Conti, and Mason (2020) Hinshaw (1992), Segal (2008), Diprete and Jennings (2011), Kristoffersen et al. (2015), Owens (2016), and Papageorge, Ronda, and Zheng (2019).
19
that children who exhibit behaviors that deviate from gender norms and stereotypes may be
penalized in the labor market.19
In our setting, dependent behavior is more prevalent among girls than boys while
headstrong behavior (along with other child behaviors) is more prevalent among boys than girls.
At the same time, numerous public opinion surveys suggest that certain traits, such as
stubbornness and decisiveness, are more associated with males while other behaviors, such as
sensitivity and being emotional, are more associated with females (Eagly et al. 2019). This
hypothesis is consistent with the role congruity theory of prejudice, which posits that men and
women who behave in ways that are contrary to expected behaviors are often subject to prejudice
(Eagly and Karau, 2002; Eagly and Koenig, 2008).20 Indeed, men and women who do not
conform to gender norms and stereotypes in the labor market have been shown to suffer social
and economic sanctions (Akerlof and Kranton 2000; Rudman and Glick, 2001; Brescoll and
Uhlmann, 2008).
Conclusion
This article adds to the literature on the long-term effects of several distinct child
behaviors. We contribute to this literature by describing gender differences in the associations of
child behaviors with earnings for a recent national cohort of children in the U.S. Our results are
novel. We find that two child behaviors, headstrong and dependent behaviors, have significantly
different associations with early adult earnings by gender. The gender differences are large. The
labor market returns to a 1σ increase in headstrong behavior during childhood differed by -$2431
19 Gender norms have also been shown to affect female attachment to paid work, and the choice of field of study and occupation (e.g., Senik and Friedman-Sokuler, 2020). 20 Role congruity theory may also be applicable to prejudice against those having characteristics that deviate from gender norms, such as physical appearance (Blakemore, 2003) or names (Figlio, 2007).
20
(female minus male), with women characterized as headstrong suffering large earnings penalties.
This figure represents approximately 40% of the gap between male and female average earnings.
At the same time, the labor market returns to a 1σ increase in dependent behavior during
childhood differed by $2036 (female minus male), with men characterized as headstrong
suffering large earnings penalties. This figure represents 36% of the male-female earnings gap.
We have been careful throughout this article to refer to the relationships between child
behaviors and adult outcomes as associations. They do not necessarily reflect causal
relationships. However, our measures of child behavior are measured early in life, and our results
are similar whether we include or omit academic achievement and family background
characteristics, and whether we allow or constrain family background characteristics to have the
same effects by gender. Therefore, while we cannot rule out the possibility that other
unmeasured factors are confounding our estimates, the evidence presented above suggests that
this may not be a serious problem. It is also important to note that we use earnings measured
relatively early in adult lives and may therefore not fully capture lifecycle earnings (even though
we control for age).
Our novel findings contribute to the growing literature on the potential causes of the
gender wage gap (Blau and Khan 2017).21 Other studies have shown that gender differences in
adult personality and the returns to those traits can explain a significant part of the gender wage
gap (Filer 1983; Nyhus and Pons, 2005; Braakmann 2009; Heineck 2011; Mueller and Plug,
2006).22 We focus on specific child behaviors, which have not been previously examined, and
21 While we show large gender differences in associations between headstrong and dependent behavior and earnings, these differences can explain only a small amount of the gender gap in earnings because of relatively small differences between genders in these behaviors. 22 A related literature explores whether the gender wage gap is due to gender differences in negotiations (Babcock and Laschever, 2003), educational choices (Zafar, 2013), or because women shy away from competitions (Niederle and Vesterlund, 2007),
21
how these behaviors affect early adult earnings. Moreover, we provide a set of facts to motivate
future research on the underlying mechanisms for the gender differences that we do (and do not)
observe. Notably, our analyses rule out several mediating pathways. It does not appear that the
gender differences in the associations between these child behaviors and earnings are explained
by education, marriage, depression, self-esteem, health, occupation, or adult personality traits.
Moreover, our measures of child behavior are not highly correlated with the adult personality
traits, suggesting that they reflect distinct attributes that affect earnings.
However, we do find that gender differences in the labor market returns to these
behaviors differ markedly by education and occupation, suggesting that the workplace, and
perhaps workplace prejudice, could be a potential explanation. We acknowledge that more
research is necessary to distinguish this hypothesis from alternative explanations. For example,
do the moderating effects of gender arise because of differences in actual productivity, or is it
because colleagues and supervisors have negative perceptions of headstrong women and
dependent men? The former could arise if behavior is more pronounced by gender while the
latter is consistent with a bias due to non-conforming gender behaviors (Eagly and Karau, 2002;
Eagly, 2004, Eagly and Koenig, 2008).
Another important question is why other child behaviors, such as hyperactive, anti-social,
and peer conflict, do not differ in their associations with earnings by gender? While not as
strongly correlated with earnings as headstrong behavior, these behaviors are more prevalent
among boys than girls and significantly correlated with adult earnings in some specifications. So
why is the moderating role of gender absent in the case of these behaviors? Perhaps these
behaviors are less associated with adult gender stereotypes and therefore not perceived as gender
non-conforming behaviors. Or perhaps these behaviors do not lead to negative perceptions on the
22
job because they do not affect social interactions. Further research is needed to answer these
questions as well.
Finally, as noted earlier, the child behaviors we examine are well defined in the child
developmental literature, and there is a large literature discussing the biological, social and
economic determinants of these and related behaviors. Children with greater behavioral problems
tend to come from poorer families and behaviors are strongly associated with parenting style
(Campbell 1995; Huaqing and Kaiser 2003). While we do not know to what extent these specific
child behaviors are malleable and, if so, at what ages, there is evidence that other child
behavioral characteristics can be altered through interventions (such as the Becoming a Man
program in Chicago) so it seems plausible that headstrong and dependent behavior are at least
partially malleable and may be affected by policy.
23
References 1. Achenbach, Thomas, and Craig Edelbrock. 1981. Behavioral problems and competencies
reported by parents of normal and disturbed children aged four through sixteen. Monographs of the Society for Research in Child Development, 46(1): 1-82.
2. Akerlof, George, and Rachel Kranton. 2000. Economics and Identity. The Quarterly Journal
of Economics, 115(3): 715-753. 3. Attanasio, Orazio, Richard Blundell, Gabriella Conti, and Giacomo Mason. 2020. Inequality
in socio-emotional skills: A cross-cohort comparison. Journal of Public Economics, 191: article 104171.
4. Autor, David, David Figlio, Krzysztof Karbownik, Jeffrey Roth, and Melanie Wasserman.
2019. Family disadvantage and the gender gap in behavioral and educational outcomes. American Economic Journal: Applied Economics, 11(3): 338-381.
5. Blau, Francine, and Lawrence Kahn. 2017. The Gender Wage Gap: Extent, Trends, and
Explanations. Journal of Economic Literature, 55(3): 789–865. 6. Braakmann, Nils. 2009. The Role of Psychological Traits for the Gender Gap in Full-Time
Employment and Wages: Evidence from Germany. SOEP papers on Multidisciplinary Panel Data Research 162, DIW Berlin, The German Socio-Economic Panel (SOEP).
7. Brescoll, Victoria, and Eric Uhlmann. 2008. Can an angry woman get ahead? Status
conferral, gender, and expression of emotion in the workplace. Psychological Science, 19(3): 268 –275.
8. Campbell, Susan. 1995. Behavior Problems in Preschool Children: A Review of Recent
Research. Journal of Child Psychology and Psychiatry, 36(1): 113-149.
9. Campbell, Susan. 1997. Behavior problems in preschool children. In Advances in clinical child psychology (pp. 1-26). Springer, Boston, MA.
10. Caspi, Avshalom, Daryl Bem, and Glen Elder Jr. 1989. Continuities and consequences of
interactional styles across the life course. Journal of Personality, 57(2): 375-406.
11. Caspi, Avshalom, and Phil Silva. 1995. Temperamental Qualities at Age Three Predict Personality Traits in Young Adulthood: Longitudinal Evidence from a Birth Cohort. Child Development, 66(2): 486-498. doi:10.2307/1131592.
12. Caspi, Avshalom, Brent Roberts, and Rebecca Shiner. 2005. Personality Development:
Stability and Change. Annual Review of Psychology, 56(1): 453–84.
24
13. Currie, Janet, and Mark Stabile. 2006. Child mental health and human capital accumulation:
The case of ADHD. Journal of Health Economics, 25(6): 1094–111.
14. DiPrete, Thomas, and Jennifer Jennings. 2011. Social and Behavioral Skills and the Gender Gap in Early Educational Achievement. Social Science Research 41(1):1-15.
15. Donnellan, Brent, Patrick Hill, and Brent Roberts. 2015. Personality development across the
life span: Current findings and future directions. In M. Mikulincer, P. R. Shaver, M. L. Cooper, and R. J. Larsen (Eds.), APA handbook of personality and social psychology, Vol. 4. Personality processes and individual differences (p. 107–126). American Psychological Association.
16. Duncan, Greg, and Katherine Magnuson. 2011. The Nature and Impact of Early Achievement Skills, Attention Skills, and Behavior Problems." In Whither Opportunity? Rising Inequality, Schools, and Children's Life Chances. Russell Sage Foundation, 47-69.
17. Eagly, Alice, and Steven Karau. 2002. Role congruity theory of prejudice toward female
leaders. Psychological Review, 109(3): 573–598. 18. Eagly, Alice, and Anne Koenig. 2008. Gender prejudice: On the risks of occupying
incongruent roles. In E. Borgida and S. T. Fiske (Eds.), Beyond common sense: Psychological science in the courtroom (pp. 63-81). Malden: Blackwell Publishing.
19. Eagly, Alice, Christa Nater, David Miller, Michèle Kaufmann, and Sabine Sczesny. 2019.
Gender stereotypes have changed: A cross-temporal meta-analysis of U.S. public opinion polls from 1946 to 2018. American Psychologist, 75(3): 301-315.
20. Eccles, Jacquelynne, Janis Jacobs, and Rena Harold. 1990. Gender role stereotypes,
expectancy effects, and parents' socialization of gender differences. Journal of Social Issues: 46(2): 183-201.
21. Ehrler, David, Gary Evans, and Ron McGhee. 1999. Extending Big-Five Theory into
Childhood: A Preliminary Investigation into the Relationship between Big-Five Personality Traits and Behavior Problems in Children. Psychology in the Schools, 36(6): 451-58.
22. Else-Quest, Nicole, Janet Hyde, Hill Goldsmith, Carol Van Hulle. 2006. Gender differences
in temperament: a meta-analysis. Psychology Bulletin, 132(1): 33-72.
23. Figlio, David. 2003. Boys Named Sue: Disruptive Children and Their Peers. Education Finance and Policy, 2(4): 376-394.
25
24. Filer, Randall K. 1983. Sexual Differences in Earnings: The Role of Individual Personalities and Tastes. The Journal of Human Resources 18(1): 82-99.
25. Fletcher, Jason, and Barbara Wolfe. 2008. Child mental health and human capital accumulation: the case of ADHD revisited. Journal of Health Economics, 27(3): 794–800.
26. Fletcher, Jason, and Barbara Wolfe. 2009. Long term consequences of childhood ADHD on criminal activities. The Journal of Mental Health Policy and Economics, 12(3): 119–138.
27. Fletcher, Jason. 2013. The effects of personality traits on adult labor market outcomes:
Evidence from siblings. Journal of Economic Behavior and Organization, 2013(89): 122-135.
28. Fletcher, Jason. 2014. The effects of childhood ADHD on adult labor market outcomes.
Health Economics, 23(2): 159-181.
29. Friedman-Sokuler, Naomi, and Claudia Senik. 2020. From Pink-Collar to Lab Coat: Cultural Persistence and Diffusion of Socialist Gender Norms. IZA Discussion Paper No. 13385.
30. Gensowski, Miriam. 2018. Personality, IQ, and lifetime earnings. Labour Economics,
51: 170-18.
31. Guowei Li, Monica Taljaard, Edwin Van den Heuvel, Mitchell Levine, Deborah Cook, George Wells, Philip Devereaux, and Lehana Thabane. 2017. An introduction to multiplicity issues in clinical trials: the what, why, when and how. International Journal of Epidemiology, 46(2): 746–755.
32. Heckman, James, Jora Stixrud, and Sergio Urzua. 2006. The effects of cognitive and non-
cognitive abilities on labor market outcomes and social behavior. Journal of Labor economics, 24(3): 411–482.
33. Heineck, Guido, and Silke Anger. 2010. The returns to cognitive abilities and personality
traits in Germany. Labour Economics, 17(3): 535–546. 34. Heineck, Giudo. 2011 Does it pay to be nice? Personality and earnings in the UK. Industrial
and Labor Relations Review, 64(5): 1020-1038.
35. Hinshaw, Stephen. 1992. Externalizing behavior problems and academic underachievement in childhood and adolescence: causal relationships and underlying mechanisms. Psychological Bulletin, 111(1): 127-155.
26
36. Huaqing Qi, Cathy, and Ann Kaiser. 2003. Behavior Problems of Preschool Children From Low-Income Families: Review of the Literature. Topics in Early Childhood Special Education, 23(4): 188-216.
37. Hyde, Janet, and Janet Mertz. 2009. Gender, culture and mathematics performance.
Proceedings of the National Academy of Sciences, 106(22): 8801-8807. 38. Judge, Timothy, Beth Livingston, and Charlice Hurst. 2012. Do Nice Guys—and Gals—
Really Finish Last? The Joint Effects of Sex and Agreeableness on Income. Journal of Personality and Social Psychology, 102(2): 390–407.
39. Kristoffersen, Jannie, Carsten Obel, and Nina Smith. 2015. Gender differences in behavioral
problems and school outcomes. Journal of Economic Behavior and Organization, 115:75-93. 40. McCrae, Robert, Paul Costa, Fritz Ostendorf, Alois Angleitner, Martina Hřebíčková, Maria
Avia, Jesús San, Maria Sánchez-Bernardos, Ersin, Kusdil, Ruth Woodfield, Peter Saunders, and Peter Smith. 2000. Nature over nurture: Temperament, personality, and life span development. Journal of Personality and Social Psychology, 78(1): 173-186.
41. McLeod, Jane, and Karen Kaiser. 2004. Childhood Emotional and Behavioral Problems and
Educational Attainment. American Sociological Review, 69(5): 636-658. 42. Mueller, Gerrit, and Erik Plug. 2006. Estimating the effect of personality on male and female
earnings. Industrial and Labor Relations Review, 60(1): 3–22. 43. Murnane, Richard, John Willett, and Frank Levy. 1995. The growing importance of cognitive
skills in wage determination. The review of economics and statistics, 77(2): 251–266. 44. Newman, Denise, Avshalom Caspi, Terrie Moffitt, and Phil Silva. 1997. Antecedents of
Adult Interpersonal Functioning: Effects of Individual Differences in Age 3 Temperament. Developmental Psychology, 33(2): 206–17.
45. Nyhus, Ellen, and Empar Pons. 2005. The effects of personality on earnings. Journal of
Economic Psychology, 26(3): 363–84.
46. Nyhus, Ellen, and Empar Pons. 2012. Personality and the gender wage gap. Applied Economics, 44(1): 105–18.
47. Owens, Jayanti. 2016. Early Childhood Behavior Problems and the Gender Gap in
Educational Attainment in the United States. Sociology of Education, 89(3): 236-258. 48. Papageorge, Nicholas, Victor Ronda, and Yu Zheng. 2019. The Economic Value of Breaking
Bad: Misbehavior, Schooling and the Labor Market. NBER Working Paper No. 25602.
27
49. Parcel, Toby, and Elizabeth Menaghan. 1988. Measuring Behavioral Problems in a Large Cross-Sectional Survey: Reliability and Validity for Children of the NLS Youth. Columbus: The Ohio State University, Department of Sociology.
50. Peterson, James, and Nicholas Zill. 1986. Marital disruption, parent–child relationships, and
behavior problems in children. Journal of Marriage and the Family, 48(2), 295-307.
51. Prinz, Daniel, Michael Chernew, David Cutler, and Austin Frakt. 2018. Health and Economic Activity Over the Lifecycle: Literature Review. NBER Working Paper No. 24865.
52. Proschan, Michael, and Myron Waclawiw, 2000. Practical Guidelines for Multiplicity
Adjustment in Clinical Trials. Controlled Clinical Trials, 21(6): 527-539. 53. Reuben, Ernesto, Paola Sapienza, and Luigi Zingales. 2014. How stereotypes impair
women’s careers in science. Proceedings of the National Academy of Sciences, 111(12): 4403-4408.
54. Roberts Brent, Avshalom Caspi, and Terrie Moffitt. 2003. Work experiences and personality
development in young adulthood. Journal of Personality and Social Psychology. 84(3): 582–93.
55. Rudman, Laurie, and Peter Glick. 2001. Prescriptive gender stereotypes and backlash toward
agentic women. Journal of Social Issues, 57(4): 743–762. 56. Segal, Carmit. 2008. Classroom Behavior. Journal of Human Resources, 43(4): 783-814. 57. Shiner, Rebecca, and Colin DeYoung. 2013. The structure of temperament and personality
traits: A developmental perspective. In P. Zelazo (Ed.), Oxford handbook of developmental psychology (pp. 113-141). New York: Oxford University Press.
58. Spencer, Steven, Claude Steele, and Diane Quinn. 1999. Stereotype threat and women's math performance. Journal of Experimental Social Psychology, 35(1): 4-28.
59. Wason, James, Lynne Stecher, and Adrian Mander. 2014. Correcting for multiple-testing in
multi-arm trials: is it necessary and is it done? Trials, 15(1): 1-7.
60. Westfall, Peter, and Stanley Young. 1993. Resampling-Based Multiple Testing: Examples and Methods for p-Value Adjustment. New York: John Wiley.
61. Yarnell, Lisa, Marsha Sargeant, Carol Prescott, Jacqueline Tilley, Jo Farver, Sarnoff
Mednick, Peter Venables, Adrian Raine, and Susan Luczak. 2013. Measurement invariance of internalizing and externalizing behavioral syndrome factors in a non‐Western sample. Assessment, 20(5): 642-55.
28
62. Zill, Nicholas. 1985. Behavior problem scales developed from the 1981 Child Health
Supplement to the National Health Interview Survey. Washington, DC: Child Trends. Retrieved from: http://www.childtrends.org/Files/Child_Trends-1990_11_01_ES_BehaviorProb.pdf
63. Zill, Nicholas. 1990. Behavior problems based on parent report. Washington, DC: Child.
Retrieved from: http://www.childtrends.org/Files/Child_Trends-1990_11_01_ES_BehaviorProb.pdf
29
Table 1: Sample Means/Proportions of Child and Mother Characteristics by Gender
Child Characteristics
Male
Female
Female-Male
Difference
P-value
Standardized Difference
Non-Hispanic Black 0.31 0.35 0.05 0.01 0.07 Non-Hispanic White 0.46 0.43 -0.02 0.24 -0.03 Hispanic 0.24 0.21 0.03 0.10 -0.04 Birth Year 1984.6 1984.4 0.19 0.02 -0.05 Birth Order 1.86 1.84 -0.02 0.60 0.01 Age at time of Adult Interview 26.4 26.5 -0.08 0.01 0.03 Behavioral Problem Subscales (avg. of ages 4-12) Dependent 1.30 1.42 0.12 0.001 0.09 Head Strong 2.50 2.22 -0.27 0.001 -0.15 Anti-Social 1.70 1.26 -0.44 0.001 -0.26 Anxious 1.62 1.66 0.04 0.370 0.02 Hyperactive 2.16 1.66 -0.50 0.001 -0.29 Peer Conflict 0.55 0.46 -0.09 0.001 -0.10 PIAT Achievement Test Scores (avg. of ages 5-13) Math “Raw” Score 36.1 36.0 -0.08 0.800 -0.01 Reading Comprehension “Raw” Score 35.1 36.5 1.49 0.001 0.11 Mother’s/Family’s Characteristics Mother’s Age at Birth 23.6 23.3 -0.32 0.005 -0.07 Mother <12 Years of Completed Education 0.24 0.23 0.00 0.76 -0.01 Mother=12 Years of Completed Education 0.41 0.40 -0.01 0.47 -0.02 Mother 13-15 Years of Completed Education 0.23 0.26 -0.03 0.13 0.04 Mother 16 Years or More of Completed Education 0.11 0.11 0.00 0.68 -0.01 Mother’s AFQT Percentile Score in 1981 35.4 34.3 -1.09 0.26 -0.03 Mother Never Married in Year of Birth 0.27 0.32 0.05 0.005 0.07 Mother Separated Divorced in Year of Birth 0.08 0.08 0.00 0.790 -0.01 Mother Married in Year of Birth 0.65 0.61 -0.05 0.010 -0.07 Mother’s Locus of Control Score in 1979 8.9 9.0 0.13 0.14 0.04 Mother’s Self-esteem Score in 1980 21.9 21.8 -0.08 0.59 -0.01 Child’s Household had Library Card 0.66 0.71 0.05 0.001 0.08 Child’s Household had Magazines 0.50 0.50 -0.00 0.910 -0.00 Child’s Household had Newspapers 0.70 0.71 0.01 0.750 0.01 Number of Observations 3557 3726 7283
: Sample includes children born between 1981 and 1990 and observed as an adult between ages of 24 to 30. Where applicable, column sums may not add to 100% because of rounding. Similarly, differences between males and females may not be exact due to rounding. The standardized difference is calculated using: [mean(M)-mean(F)]/[sqrt(var(M)+var(F))]. ** p-value<=0.05
30
Table 2: Estimates of Associations between Child Behavior and Adult Earned Income
Annual Earnings Natural Logarithm Earnings (Eaarnings>0)
Males Females Female-Male Males Females Female-Male Dependent -1632** 431 2063** -0.085** 0.036 0.121** (662) (544) (856) (0.027) (0.026) (0.038) Headstrong 339 -2092** -2431** 0.035 -0.099** -0.134** (926) (710) (1167) (0.036) (0.034) (0.050) Antisocial -1034 -123 911 -0.082** -0.023 0.059 (765) (679) (1023) (0.033) (0.040) (0.052) Anxious/Depressed 1313 210 -1103 0.052 0.053* 0.001 (808) (660) (1043) (0.034) (0.031) (0.046) Hyperactive -1036 -925 111 -0.030 -0.066* -0.037 (793) (636) (1016) (0.033) (0.034) (0.048) Peer Conflict -1308** -909 399 -0.053* -0.046 0.007 (659) (573) (874) (0.031) (0.034) (0.046) Mean of Dep. Variable 26746 20847 10.02 9.81 (standard deviation) (24388) (19628) (0.97) (0.98) Number of observations 3557 3726 7283 3053 3054 6107
Notes: Sample includes children born between 1981 and 1990 and observed as an adult between ages of 24 to 30. All regression models include dummy variables for child’s race (white, black, Hispanic), each year of age, each birth year and each birth order, and quartiles of PIAT math and reading scores. Regressions also include mother’s characteristics: dummy variables for each age at birth; dummy variables for education (LTHS, HS, some college, BA or more); AFQT score and its square; marital status at birth (married, never married, other); dummy variables for quartile of self-esteem scale; dummy variables for quartile of Rotter scale; and dummy variables indicating whether household of child had magazines, newspapers and a library card. Child behaviors are measured in standard deviation units. Robust standard errors clustered within person in parentheses. * 0.10<=p-value<0.05, ** p-value<=0.05.
31
Table 3: Estimates of Associations between Child Behavior and Employment, Education, Marital Status, Health and Number of Children
Employed at
Time of Interview
Hours of Work Week of Interview (Hours>0)
Years of Completed Education
Bachelor’s Degree of More
Married Poor Health Good Health Number of Children
Female-Male
Female-Male
Female-Male
Female-Male
Female-Male
Female-Male
Female-Male
Female-Male
Dependent 0.020 0.536 0.161* 0.002 -0.024 -0.027** 0.001 -0.006 (0.017) (0.540) (0.094) (0.013) (0.018) (0.013) (0.015) (0.049) Headstrong -0.034* 0.730 -0.066 -0.011 -0.007 0.027* -0.001 0.067 (0.021) (0.676) (0.123) (0.018) (0.024) (0.016) (0.019) (0.060) Antisocial 0.032 -1.305** 0.083 -0.001 0.025 -0.012 0.028 0.004 (0.021) (0.649) (0.121) (0.016) (0.023) (0.018) (0.019) (0.065) Anxious/Depressed -0.005 -0.712 -0.092 -0.018 0.011 -0.011 -0.046** 0.047 (0.019) (0.682) (0.113) (0.015) (0.022) (0.016) (0.018) (0.058) Hyperactive 0.009 -0.136 0.021 0.016 -0.021 0.000 0.011 -0.026 (0.021) (0.613) (0.115) (0.015) (0.021) (0.016) (0.018) (0.057) Peer Conflict -0.007 0.421 -0.094 -0.011 -0.013 0.024 -0.002 0.056 (0.019) (0.575) (0.104) (0.014) (0.020) (0.016) (0.015) (0.053) Mean of Dep. Variable 0.72 40.7 11.1 0.15 0.26 0.14 0.20 0.87 (standard deviation) (12.5) (4.0) Number of Obs. 7283 5194 7283 7283 7283 7283 7283 7283
Notes: Sample includes children born between 1981 and 1990 and observed as an adult between ages of 24 to 30. All regression models include dummy variables for child’s race (white, black, Hispanic), each year of age, each birth year and each birth order, and quartiles of PIAT math and reading scores. Regressions also include mother’s characteristics: dummy variables for each age at birth; dummy variables for education (LTHS, HS, some college, BA or more); AFQT score and its square; marital status at birth (married, never married, other); dummy variables for quartile of self-esteem scale; dummy variables for quartile of Rotter scale; and dummy variables indicating whether household of child had magazines, newspapers and a library card. Child behaviors are measured in standard deviation units. Robust standard errors clustered within person in parentheses. * 0.10<=p-value<0.05, ** p-value<=0.05.
32
Table 4: Estimates of Gender Differences in Association between Child Behavior, Math and Reading Achievement and Depression, Mastery, Self-Esteem and Personality
Depression (CESD)
Pearlin Mastery
Self-esteem Extroversion
Openness Agreeableness Conscientiousness Emotional Stability
Female-Male Female-Male Female-Male Female-Male Female-Male Female-Male Female-Male Female-Male Dependent -0.035 -0.109 0.340* 0.048 -0.001 0.031 -0.000 -0.039 (0.178) (0.129) (0.199) (0.062) (0.050) (0.050) (0.051) (0.059) Headstrong 0.110 -0.115 0.123 -0.079 0.045 -0.063 -0.034 0.072 (0.224) (0.163) (0.253) (0.078) (0.062) (0.064) (0.060) (0.071) Antisocial 0.032 0.072 -0.216 0.163** -0.024 0.040 -0.001 -0.051 (0.231) (0.166) (0.255) (0.078) (0.061) (0.063) (0.062) (0.075) Anxious/Depressed 0.155 0.101 -0.229 0.088 -0.055 0.025 -0.065 -0.092 (0.221) (0.157) (0.242) (0.073) (0.060) (0.059) (0.059) (0.069) Hyperactive 0.255 0.049 -0.107 -0.059 -0.006 -0.084 0.025 0.013 (0.221) (0.149) (0.234) (0.075) (0.059) (0.060) (0.058) (0.066) Peer Conflict -0.358 0.007 -0.020 -0.089 -0.018 -0.017 0.002 0.129** (0.218) (0.160) (0.240) (0.071) (0.054) (0.057) (0.055) (0.064) Mean of Dep. Variable 4.7 13.3 32.7 4.6 5.4 5.1 5.8 5.1 (standard deviation) (3.8) (2.8) (4.2) (1.4) (1.1) (1.2) (1.1) (1.3) Number of Observations 3128 3128 3128 4103 4121 4070 4125 4120
Notes: Sample includes children born between 1981 and 1990. For CESD, Pearlin Mastery and Self-esteem measures, respondents are between ages 18 and 23 because information on Pearlin and Self-esteem was not obtained in all survey years (available in 1998 and 2004). For measures of personality traits, information was available in 2006, 2010 and 2014 and respondent is observed as an adult between ages of 24 to 30. All regression models include dummy variables for child’s race (white, black, Hispanic), each year of age, each birth year and each birth order, and quartiles of PIAT math and reading scores. Regressions also include mother’s characteristics: dummy variables for each age at birth; dummy variables for education (LTHS, HS, some college, BA or more); AFQT score and its square; marital status at birth (married, never married, other); dummy variables for quartile of self-esteem scale; dummy variables for quartile of Rotter scale; and dummy variables indicating whether household of child had magazines, newspapers and a library card. Child behaviors are measured in standard deviation units. Robust standard errors clustered within person in parentheses. * 0.10<=p-value<0.05, ** p-value<=0.05.
Table 5: Mediation Analysis Estimates of Associations between Child Behavior and Adult Earned Income Controlling for Potential Mediators
Estimate from
Primary Analysis Restricted to Sample with Occupation
Include Occupation
Include Occupation and Education
Include Additional Mediators*
Female-Male Female-Male Female-Male Female-Male Dependent 2063** 1982** 1409* 1382* 1201 (856) (880) (814) (810) (800) Headstrong -2431** -1963 -2323** -2214** -2074* (1167) (1200) (1083) (1068) (1059) Antisocial 911 886 1045 1208 1513 (1023) (1066) (978) (967) (947) Anxious/Depressed -1103 -1429 -552 -598 -546 (1043) (1081) (990) (982) (962) Hyperactive 111 -363 56 121 104 (1016) (1037) (954) (946) (932) Peer Conflict 399 327 203 315 64 (874) (895) (837) (833) (821) Mean of Dep. Variable 26746 26103 26103 26103 26103 (standard deviation) (24388) (22030) (22030) (22030) (22030) Number of observations 7283 6469 6469 6469 6469
Notes: Sample includes children born between 1981 and 1990 and observed as an adult between ages of 24 to 30. Unless indicated, all regression models include dummy variables for child’s race (white, black, Hispanic), each year of age, each birth year and each birth order, and quartiles of PIAT math and reading scores. Regressions also include mother’s characteristics: dummy variables for each age at birth; dummy variables for education (LTHS, HS, some college, BA or more); AFQT score and its square; marital status at birth (married, never married, other); dummy variables for quartile of self-esteem scale; dummy variables for quartile of Rotter scale; and dummy variables indicating whether household of child had magazines, newspapers and a library card. Child behaviors are measured in standard deviation units. Robust standard errors clustered within person in parentheses. * 0.10<=p-value<0.05, ** p-value<=0.05. *Mediators include: occupation dummy variables (2-digit), education dummy variables (LTHS, HS, SC, BA), Marital Status (married, never married, other), dummy variables for number of children, and indicator of poor health.
Table 6: Heterogeneity (Moderation) Analysis Estimates of Associations between Child Behavior and Adult Earned Income by Education and Occupational Characteristics
Estimates
Primary Analysis
<HS >=HS Blue Collar Not Blue Collar
>=75% Female
Occupation
<75% Female
Occupation
Female-Male Female-Male
Female-Male
Female-Male
Female-Male
Female-Male Female-Male
Dependent 2063** 3198** 899 2399** 1713 2622 2079* (856) (1094) (1080) (1023) (1383) (1662) (1090) Headstrong -2431** -3151** -1498 -2715* -625 -1772 -2543* (1167) (1517) (1442) (1512) (1657) (1748) (1445) Antisocial 911 2282* -62 1514 -396 -3987** 2223* (1023) (1282) (1293) (1282) (1563) (1926) (1300) Anxious/Depressed -1103 -1539 -911 -732 -1854 1234 -968 (1043) (1346) (1271) (1320) (1478) (1890) (1336) Hyperactive 111 -793 472 683 -1226 -1985 -1035 (1016) (1289) (1270) (1291) (1572) (1936) (1225) Peer Conflict 399 905 -12 1459 30 1508 -162 (874) (1114) (1085) (1086) (1289) (1923) (1097) Mean of Dep. Variable
23728 22611 24424 22478 28932 20994 27587
(standard deviation) (22276) (23170) (21674) (20360) (22856) (16193) (23227) Number of observations
7283 2794 4489 2836 3633 1640 4773
Notes: Sample includes children born between 1981 and 1990 and observed as an adult between ages of 24 to 30. All regression models include dummy variables for child’s race (white, black, Hispanic), each year of age, each birth year and each birth order, and quartiles of PIAT math and reading scores. Regressions also include mother’s characteristics: dummy variables for each age at birth; dummy variables for education (LTHS, HS, some college, BA or more); AFQT score and its square; marital status at birth (married, never married, other); dummy variables for quartile of self-esteem scale; dummy variables for quartile of Rotter scale; and dummy variables indicating whether household of child had magazines, newspapers and a library card. Child behaviors are measured in standard deviation units. Robust standard errors clustered within person in parentheses. * 0.10<=p-value<0.05, ** p-value<=0.05.
Appendix Table 1: Behavior Problems Index (BPI) questions and subscales
Question Subscale (He or She) clings to adults Dependent (He or She) cries too much Dependent (He or She) demands a lot of attention Dependent (He or She) is too dependent on others Dependent (He or She) is rather high strung, tense, and nervous Headstrong (He or She) argues too much Headstrong (He or She) is disobedient at home Headstrong (He or She) is stubborn, sullen, or irritable Headstrong (He or She) has strong temper and loses it easily Headstrong (He or She) cheats or tells lies Antisocial (He or She) bullies or is cruel/mean to others Antisocial (He or She) does not seem to feel sorry after misbehaving Antisocial (He or She) breaks things deliberately Antisocial (He or She) is disobedient at school (>5 yrs) Antisocial (He or She) has trouble getting along with teachers (>5 yrs) Antisocial (He or She) has sudden changes in mood or feeling Anxious/Depressed (He or She) feels/complains no one loves him/her Anxious/Depressed (He or She) is too fearful or anxious Anxious/Depressed (He or She) feels worthless or inferior Anxious/Depressed (He or She) is unhappy, sad, or depressed Anxious/Depressed (He or She) has difficulty concentrating/paying attention Hyperactive (He or She) is easily confused, seems in a fog Hyperactive (He or She) is impulsive or acts without thinking Hyperactive (He or She) has trouble getting mind off certain thoughts Hyperactive (He or She) is restless, overly active, cannot sit still Hyperactive (He or She) has trouble getting along with other children Peer Conflict (He or She) is not liked by other children Peer Conflict (He or She) is withdrawn, does not get involved with others Peer Conflict
Appendix Table 2: Estimates of Associations between Child Behavior and Maternal characteristics
Mom Age at Birth Mom has LTHS Mom has BA Mom’s AFQT Mom Married Mom’s Locus of Control
Female-Male Female-Male Female-Male Female-Male Female-Male Female-Male
Dependent -0.199 -0.031 0.004 1.724 0.050** 0.023 (0.135) (0.019) (0.011) (1.167) (0.022) (0.114) Headstrong 0.000 0.001 -0.013 -0.637 0.002 0.053 (0.175) (0.023) (0.016) (1.466) (0.026) (0.143) Antisocial 0.053 0.006 -0.013 -2.502 -0.038 -0.041 (0.171) (0.025) (0.014) (1.487) (0.029) (0.134) Anxious/Depressed 0.255 -0.025 0.056** 3.644** -0.007 -0.134 (0.164) (0.022) (0.016) (1.506) (0.026) (0.137) Hyperactive -0.156 0.031 -0.008 -0.606 -0.009 0.080 (0.160) (0.021) (0.015) (1.396) (0.026) (0.132) Peer Conflict -0.048 0.036 -0.022 -2.294 0.001 0.054 (0.152) (0.022) (0.013) (1.342) (0.025) (0.130) Number of Obs. 6159 6159 6159 6159 6159 6159
Notes: Sample includes children born between 1981 and 1990 and observed as an adult between ages of 24 to 30. Child behaviors are measured in standard deviation units. Robust standard errors clustered within person in parentheses. * 0.10<=p-value<0.05, ** p-value<=0.05.
Appendix Table 3. Sensitivity Analysis Estimates of Associations between Child Behavior and Adult Earned Income
Estimate
Primary Analysis
Restrict Family
Background
Child Born 1981-84
Child Born 1985-90
Dependent and Head
Strong Only
No Mother or Family
Background
Female-Male
Female-Male
Female-Male
Female-Male
Female-Male
Female-Male
Dependent 2063** 1955** 1362 3491** 2004** 2066** (856) (874) (1230) (1114) (806) (877) Headstrong -2431** -2106* -2785* -2332 -2018** -2290** (1167) (1144) (1652) (1636) (819) (1159) Antisocial 911 932 1873 319 1136 (1023) (1015) (1463) (1385) (1041) Anxious/Depressed -1103 -1313 -448 -1970 -1046 (1043) (1040) (1585) (1287) (1038) Hyperactive 111 109 382 -579 -67 (1016) (1005) (1538) (1273) (1017) Peer Conflict 399 90 -507 1226 -211 (874) (883) (1250) (1150) (891) Mean of Dep. Variable 23728 23728 23695 23769 23728 23728 (standard deviation) (22276) (22276) (22479) (22033) (22276) (22276) Number of observations
7283 7283 3973 3310 7283 7283
Notes: Sample includes children born between 1981 and 1990 and observed as an adult between ages of 24 to 30. All regression models include dummy variables for child’s race (white, black, Hispanic), each year of age, each birth year and each birth order, and quartiles of PIAT math and reading scores. Regressions also include mother’s characteristics: dummy variables for each age at birth; dummy variables for education (LTHS, HS, some college, BA or more); AFQT score and its square; marital status at birth (married, never married, other); dummy variables for quartile of self-esteem scale; dummy variables for quartile of Rotter scale; and dummy variables indicating whether household of child had magazines, newspapers and a library card. Child behaviors are measured in standard deviation units. Robust standard errors clustered within person in parentheses. * 0.10<=p-value<0.05, ** p-value<=0.05.
Appendix Table 4: Estimates of Associations between Child Behavior and Employment, Education, Marital Status, Health and Number of Children
Employed at Time of Interview Years of Education Bachelor’s Degree of More
Males
Females
Female-Male
Males
Females
Female-Male
Males
Females
Female-Male Dependent -0.005 0.015 0.020 0.055 0.215** 0.161* 0.007 0.009 0.002 (0.012) (0.012) (0.017) (0.068) (0.065) (0.094) (0.008) (0.010) (0.013) Headstrong -0.002 -0.036** -0.034* -0.078 -0.144* -0.066 -0.008 -0.019 -0.011 (0.014) (0.015) (0.021) (0.089) (0.084) (0.123) (0.012) (0.014) (0.018) Antisocial -0.034** -0.002 0.032 -0.273** -0.190** 0.083 -0.016* -0.017 -0.001 (0.013) (0.016) (0.021) (0.083) (0.088) (0.121) (0.010) (0.013) (0.016) Anxious/Depressed 0.023* 0.017 -0.005 0.147* 0.056 -0.092 0.013 -0.004 -0.018 (0.013) (0.014) (0.019) (0.082) (0.079) (0.113) (0.010) (0.012) (0.015) Hyperactive -0.027* -0.018 0.009 -0.235** -0.214** 0.021 -0.031** -0.015 0.016 (0.014) (0.016) (0.021) (0.078) (0.084) (0.115) (0.010) (0.012) (0.015) Peer Conflict -0.013 -0.021 -0.007 -0.022 -0.116 -0.094 -0.004 -0.016 -0.011 (0.012) (0.015) (0.019) (0.071) (0.075) (0.104) (0.009) (0.011) (0.014) Mean of Dep. Variable 0.72 0.71 10.7 11.5 0.12 0.17 (standard deviation) Number of Obs. 3557 3726 7283 3557 3726 7283 3557 3726 7283 Married Poor Health Number of Children
Males
Females
Female-Male
Males
Females
Female-Male
Males
Females
Female-Male Dependent 0.011 -0.012 -0.024 0.032** 0.005 -0.027** -0.028 -0.034 -0.006 (0.012) (0.013) (0.018) (0.009) (0.010) (0.013) (0.033) (0.036) (0.049) Headstrong -0.007 -0.014 -0.007 -0.017 0.010 0.027* -0.013 0.054 0.067 (0.016) (0.018) (0.024) (0.011) (0.012) (0.016) (0.039) (0.046) (0.060) Antisocial 0.000 0.025 0.025 -0.002 -0.014 -0.012 0.145** 0.149** 0.004 (0.014) (0.018) (0.023) (0.011) (0.014) (0.018) (0.037) (0.053) (0.065) Anxious/Depressed 0.008 0.019 0.011 0.016 0.005 -0.011 -0.045 0.002 0.047 (0.014) (0.017) (0.022) (0.012) (0.011) (0.016) (0.038) (0.043) (0.058) Hyperactive -0.002 -0.023 -0.021 0.012 0.012 0.000 0.007 -0.019 -0.026 (0.014) (0.015) (0.021) (0.011) (0.012) (0.016) (0.033) (0.046) (0.057) Peer Conflict -0.020 -0.032** -0.013 -0.000 0.023* 0.024 -0.088** -0.032 0.056 (0.013) (0.016) (0.020) (0.011) (0.012) (0.016) (0.031) (0.043) (0.053) Mean of Dep. Variable 0.23 0.29 0.13 0.15 0.67 1.1 (standard deviation) Number of Obs. 3557 3726 7283 3557 3726 7283 3557 3726 7283
See notes to Table 3 in text.
Appendix Table 5: Estimates of Associations between Child Behavior, Math and Reading Achievement and Education, Marital Status, and Health
Depression (CESD) Pearlin Mastery Self-Esteem
Males
Females
Female-Male
Males
Females
Female-Male
Males
Females
Female-Male Dependent -0.047 -0.082 -0.035 0.021 -0.088 -0.109 -0.040 0.301** 0.340* (0.117) (0.134) (0.178) (0.093) (0.090) (0.129) (0.144) (0.138) (0.199) Headstrong 0.000 0.110 0.110 0.057 -0.059 -0.115 -0.035 0.088 0.123 (0.154) (0.162) (0.224) (0.116) (0.115) (0.163) (0.181) (0.176) (0.253) Antisocial -0.121 -0.089 0.032 -0.071 0.001 0.072 0.186 -0.029 -0.216 (0.139) (0.185) (0.231) (0.107) (0.127) (0.166) (0.163) (0.196) (0.255) Anxious/Depressed 0.126 0.281* 0.155 0.007 0.108 0.101 -0.027 -0.256 -0.229 (0.144) (0.168) (0.221) (0.113) (0.109) (0.157) (0.166) (0.176) (0.242) Hyperactive 0.182 0.437** 0.255 0.299** 0.348** 0.049 -0.322** -0.429** -0.107 (0.138) (0.173) (0.221) (0.104) (0.107) (0.149) (0.161) (0.171) (0.234) Peer Conflict 0.293** -0.065 -0.358 0.138 0.145 0.007 -0.318** -0.338* -0.020 (0.134) (0.172) (0.218) (0.108) (0.118) (0.160) (0.162) (0.177) (0.240) Mean of Dep. Variable 4.3 5.1 13.2 13.4 32.9 32.6 (standard deviation) (3.4) (4.0) (2.8) (2.8) (4.2) (4.3) Number of Obs. 1529 1599 3128 1529 1599 3128 1529 1599 3128 Extroversion Openness Agreeableness
Males
Females
Female-Male
Males
Females
Female-Male
Males
Females
Female-Male Dependent -0.033 0.016 0.048 -0.029 -0.030 -0.001 -0.010 0.021 0.031 (0.046) (0.041) (0.062) (0.037) (0.034) (0.050) (0.036) (0.034) (0.050) Headstrong 0.127** 0.048 -0.079 0.044 0.089** 0.045 -0.010 -0.073* -0.063 (0.055) (0.055) (0.078) (0.047) (0.041) (0.062) (0.047) (0.044) (0.064) Antisocial 0.055 0.218** 0.163** 0.084** 0.061 -0.024 -0.059 -0.019 0.040 (0.052) (0.058) (0.078) (0.040) (0.046) (0.061) (0.041) (0.048) (0.063) Anxious/Depressed -0.134** -0.046 0.088 -0.058 -0.113** -0.055 -0.034 -0.009 0.025 (0.052) (0.051) (0.073) (0.045) (0.040) (0.060) (0.041) (0.043) (0.059) Hyperactive 0.003 -0.055 -0.059 -0.039 -0.045 -0.006 0.053 -0.031 -0.084 (0.053) (0.053) (0.075) (0.042) (0.042) (0.059) (0.042) (0.042) (0.060) Peer Conflict -0.096** -0.184** -0.089 0.012 -0.006 -0.018 0.025 0.009 -0.017 (0.047) (0.053) (0.071) (0.037) (0.039) (0.054) (0.038) (0.043) (0.057) Mean of Dep. Variable 4.5 4.7 5.4 5.4 4.8 5.3 (standard deviation) (1.4)) (1.4) (1.2) (1.1) (1.2) (1.1) Number of Obs. 2028 2075 4103 2035 2086 4121 2011 2059 4070
See notes to Table 4 in text.
Appendix Table 5, Continued
Conscientiousness Emotional Stability
Males
Females Female-
Male
Males
Females Female-
Male Dependent -0.001 -0.001 -0.000 0.002 -0.037 -0.039 (0.036) (0.036) (0.051) (0.042) (0.041) (0.059) Headstrong 0.067 0.034 -0.034 -0.077 -0.005 0.072 (0.045) (0.040) (0.060) (0.049) (0.052) (0.071) Antisocial 0.031 0.030 -0.001 0.059 0.008 -0.051 (0.040) (0.047) (0.062) (0.045) (0.059) (0.075) Anxious/Depressed 0.007 -0.058 -0.065 -0.012 -0.104** -0.092 (0.042) (0.041) (0.059) (0.047) (0.050) (0.069) Hyperactive -0.168** -0.143** 0.025 -0.029 -0.015 0.013 (0.040) (0.042) (0.058) (0.044) (0.050) (0.066) Peer Conflict -0.010 -0.008 0.002 -0.155** -0.026 0.129** (0.037) (0.041) (0.055) (0.041) (0.049) (0.064) Mean of Dependent Variable 5.7 5.9 5.3 5.0 (standard deviation) (1.1) (1.1) (1.3) (1.3) Number of Observations 2040 2085 4125 2033 2087 4120
See notes to Table 4 in text.
Appendix Table 6: Estimates of Associations between Child Behavior and Work Styles
Achievement /Effort
Persistence Initiative Leadership Cooperation Concern for Others
Social Orientation
Self-Control
Female-Male Female-Male Female-Male Female-Male Female-Male Female-Male Female-Male Female-Male
Dependent -0.004 -0.017 0.001 0.003 -0.012 -0.016 0.003 0.002 (0.01) (0.01) (0.02) (0.01) (0.02) (0.01) (0.01) (0.01) Headstrong -0.009 0.016 0.015 -0.003 0.004 0.014 -0.002 0.008 (0.02) (0.02) (0.03) (0.01) (0.02) (0.02) (0.01) (0.01) Antisocial 0.001 -0.013 -0.015 -0.006 -0.02 -0.008 -0.004 -0.002 (0.02) (0.02) (0.02) (0.01) (0.02) (0.01) (0.01) (0.01) Anxious/Depressed -0.021 -0.021 -0.035 -0.005 -0.015 -0.014 -0.008 -0.003 (0.02) (0.02) (0.02) (0.01) (0.02) (0.01) (0.01) (0.01) Hyperactive 0.003 0.003 -0.012 -0.004 0.016 0.012 0.006 -0.012 (0.02) (0.02) (0.02) (0.01) (0.02) (0.01) (0.01) (0.01) Peer Conflict 0.021 0.019 0.013 0.019* 0.011 0.009 0.008 0.007 (0.01) (0.02) (0.02) (0.01) (0.02) (0.01) (0.01) (0.01) Number of Obs. 6159 6159 6159 6159 6159 6159 6159 6159
Notes: Sample includes children born between 1981 and 1990 and observed as an adult between ages of 24 to 30. All regression models include dummy variables for child’s race (white, black, Hispanic), each year of age, each birth year and each birth order, and quartiles of PIAT math and reading scores. Regressions also include mother’s characteristics: dummy variables for each age at birth; dummy variables for education (LTHS, HS, some college, BA or more); AFQT score and its square; marital status at birth (married, never married, other); dummy variables for quartile of self-esteem scale; dummy variables for quartile of Rotter scale; and dummy variables indicating whether household of child had magazines, newspapers and a library card. Child behaviors are measured in standard deviation units. Robust standard errors clustered within person in parentheses. * 0.10<=p-value<0.05, ** p-value<=0.05.
Appendix Table 6 (cont.): Estimates of Associations between Child Behavior and Work Styles
Stress-Tolerance Adaptability /Flexibility
Dependability Attention to Detail
Integrity Independence Innovation Analytical-Thinking
Female-Male Female-Male Female-Male Female-Male Female-Male Female-Male Female-Male Female-Male
Dependent -0.008 -0.017 -0.003 -0.024 -0.005 -0.018 -0.037** -0.021 (0.01) (0.02) (0.01) (0.02) (0.01) (0.01) (0.02) (0.02) Headstrong -0.001 0.023 0.007 0.018 -0.007 0.007 0.003 0.011 (0.02) (0.02) (0.02) (0.03) (0.02) (0.02) (0.02) (0.02) Antisocial 0.00 -0.035 -0.006 -0.022 -0.008 -0.001 -0.011 -0.002 (0.02) (0.02) (0.02) (0.03) (0.02) (0.02) (0.02) (0.02) Anxious/Depressed -0.012 -0.004 -0.025 -0.011 -0.02 -0.009 -0.007 -0.02 (0.02) (0.02) (0.02) (0.03) (0.02) (0.02) (0.02) (0.02) Hyperactive -0.009 -0.024 0.003 0.001 0.001 0.022 0.026 0.029* (0.02) (0.02) (0.02) (0.03) (0.02) (0.02) (0.02) (0.02) Peer Conflict 0.005 0.018 0.001 0.011 0.013 -0.004 0.018 0.003 (0.01) (0.02) (0.02) (0.03) (0.01) (0.02) (0.02) (0.02) Number of Obs. 6159 6159 6159 6159 6159 6159 6159 6159
Notes: Sample includes children born between 1981 and 1990 and observed as an adult between ages of 24 to 30. All regression models include dummy variables for child’s race (white, black, Hispanic), each year of age, each birth year and each birth order, and quartiles of PIAT math and reading scores. Regressions also include mother’s characteristics: dummy variables for each age at birth; dummy variables for education (LTHS, HS, some college, BA or more); AFQT score and its square; marital status at birth (married, never married, other); dummy variables for quartile of self-esteem scale; dummy variables for quartile of Rotter scale; and dummy variables indicating whether household of child had magazines, newspapers and a library card. Child behaviors are measured in standard deviation units. Robust standard errors clustered within person in parentheses. * 0.10<=p-value<0.05, ** p-value<=0.05.
Appendix Figure 1. Behavioral Problems Sub-scales by Age and Gender
Note: blue lines represent boys; orange lines represent girls.
0
0.5
1
1.5
2
4 5 6 7 8 9 10 11 12
Anti-social Behavior by Age and GenderNLSY Children From Birth Years 1981 to
1990
0
0.5
1
1.5
2
4 5 6 7 8 9 10 11 12
Anxious Behavior by Age and GenderNLSY Children From Birth Years 1981 to
1990
0
0.5
1
1.5
2
2.5
4 5 6 7 8 9 10 11 12
Hyperactive Behavior by Age and GenderNLSY Children From Birth Years 1981 to
1990
00.10.20.30.40.50.6
4 5 6 7 8 9 10 11 12
Peer Conflict Behavior by Age and GenderNLSY Children From Birth Years 1981 to
1990
0
0.5
1
1.5
2
4 5 6 7 8 9 10 11 12
Dependent Behavior by Age and GenderNLSY Children From Birth Years 1981 to
1990
0
0.5
1
1.5
2
2.5
4 5 6 7 8 9 10 11 12
Head Strong Behavior by Age and GenderNLSY Children From Birth Years 1981 to
1990